Manulife Sets Course to Become an AI-Powered Insurer
Manulife Financial Corporation is moving from AI pilots to an enterprise platform built for scale. The insurer has selected Akka to provide runtime infrastructure and AdaptiveML as its reinforcement-learning engine, with a clear mandate: embed AI across underwriting, distribution, and internal operations.
With roughly 37,000 employees and more than C$1 trillion in assets under management and administration in Manulife Wealth and Asset Management, this is about running AI as a capability, not a series of experiments. "In our refreshed enterprise strategy, driving AI value is one of our five key pillars," says Jodie Wallis, Global Chief AI Officer. "Before 2025 it was an enabler of business strategy, now we believe it is an integral part of the business strategy."
From Pilots to Enterprise Infrastructure
Manulife began investing in AI in 2016 with machine learning for propensity modeling, pricing analytics, and fraud detection. From 2016 to 2023, the company stood up about 70 use cases.
Generative AI accelerated that curve. The team deployed 140 use cases across 2024 and 2025-about 70 per year-with another 200 planned for 2026. The shift is just as important as the volume: AI is now embedded directly into high-volume workflows, not parked in offline analytics. As Wallis puts it, once AI runs inside core processes, it becomes a core operational capability.
Why Akka-and What the Platform Must Do
Before picking a partner, Manulife defined the platform capabilities required to support hundreds of AI services safely and efficiently. Three priorities stood out:
- Standardized developer experience: As more teams build AI solutions, the company needs consistent engineering practices, reusable components, and built-in governance. "We're moving from a small group of specialists to a wider group of colleagues building solutions," Wallis says.
- Operational reliability and visibility: AI now sits in customer-facing and internal workflows. "If AI is going to be part of how we deliver services to our customers and employees, then we need to treat it like any other high-availability platform-low latency, high availability and full visibility," she notes.
- Compute efficiency: Large models strain GPU capacity and budgets. Manulife aims to optimize performance, financial cost, and environmental impact, leveraging an already cloud-centric estate (80%+ of apps in the cloud).
Akka provides the runtime layer to operationalize these requirements. AdaptiveML brings reinforcement learning to optimize growth and govern use cases at scale-closing the loop between model performance, business outcomes, and policy controls.
Responsible AI at Scale
Manulife pairs speed with safeguards. The company's Responsible AI Principles guide value delivery for customers, colleagues, and society. Every AI system goes through model risk management, with testing and validation calibrated to materiality.
"We're not skimping on governance," Wallis says. "For every dollar we are investing in deploying AI solutions, we are also investing in AI safety." Techniques in production include two-model patterns (one to answer, one to validate), strict grounding to approved corporate knowledge sources, and adversarial testing against known prompt attacks.
For context, see the NIST AI Risk Management Framework (NIST AI RMF) and the OWASP Top 10 for LLM Applications (OWASP LLM Top 10).
Priority Use Cases
- Distribution: In Asia, Manulife supports about 106,000 agents and advisors. AI helps prep meetings, surface needs, and personalize outreach.
- Underwriting: AI extracts and structures information from physician statements, lab results, and other records to accelerate decisions. Manulife was the first life insurer in Canada to apply AI in underwriting and recently upgraded its electronic application and its MAUDE engine (Manulife Automated Underwriting Decision Engine).
- Claims: For health products, AI reads non-standard claim documents and pulls the details needed for automated adjudication.
- Intelligent document processing: "Any document, any time, any format"-AI makes sense of it, extracts what matters, and activates downstream automation.
AI for the Workforce
- Employee support: Virtual assistants handle HR questions, procurement requests, and IT tickets.
- AI for tech: Tools assist engineers with code, testing, and environment provisioning.
- Adoption at scale: "Prompt-a-thons" help teams re-work daily tasks with AI. In 2026, more than 70% of employees use AI tools regularly.
What This Signals for Insurers
The industry spent a decade becoming data-driven. The next step is operational: run AI as a managed capability that lives inside core processes with clear SLOs, governance, and cost controls. As Wallis says, "In the past we talked about being data-driven. Today we talk about being AI-powered."
Practical Takeaways for Carrier Leaders
- Stand up a standardized AI developer platform with reusable patterns, built-in policy, and observability.
- Treat AI like a production platform: define latency/availability targets, monitor end-to-end, and plan for failover.
- Manage GPU spend and capacity: choose smaller models when viable, use grounding and caching, and right-size inference.
- Expand governance with model risk tiers, adversarial testing, and human oversight on material decisions.
- Prioritize document-heavy workflows (underwriting, claims, service) for near-term ROI.
Further Reading
Your membership also unlocks:
